CN108540384B - Intelligent rerouting method and device based on congestion awareness in software defined network - Google Patents

Intelligent rerouting method and device based on congestion awareness in software defined network Download PDF

Info

Publication number
CN108540384B
CN108540384B CN201810332577.0A CN201810332577A CN108540384B CN 108540384 B CN108540384 B CN 108540384B CN 201810332577 A CN201810332577 A CN 201810332577A CN 108540384 B CN108540384 B CN 108540384B
Authority
CN
China
Prior art keywords
path
congestion
network
link
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201810332577.0A
Other languages
Chinese (zh)
Other versions
CN108540384A (en
Inventor
曲桦
赵季红
仝梦菲
陈梁骏
赵建龙
李彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN201810332577.0A priority Critical patent/CN108540384B/en
Publication of CN108540384A publication Critical patent/CN108540384A/en
Application granted granted Critical
Publication of CN108540384B publication Critical patent/CN108540384B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/12—Shortest path evaluation
    • H04L45/124—Shortest path evaluation using a combination of metrics
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/12—Shortest path evaluation
    • H04L45/121—Shortest path evaluation by minimising delays
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/12—Shortest path evaluation
    • H04L45/123—Evaluation of link metrics
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/12—Shortest path evaluation
    • H04L45/125—Shortest path evaluation based on throughput or bandwidth
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/302—Route determination based on requested QoS
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L45/00—Routing or path finding of packets in data switching networks
    • H04L45/70—Routing based on monitoring results

Abstract

A network monitoring module of a control layer regularly collects and calculates the network state of a data layer after initialization, a congestion detection module of the control layer performs fuzzy evaluation on two indexes of a used load ratio and a load change rate of a link to obtain a fuzzy evaluation value of the congestion quality of a current link, and the fuzzy evaluation value of the congestion quality of the current link is obtained according to the average congestion degree of all links on the path; after the congestion detection module of the control layer finishes the congestion quality evaluation of all the alternative paths, the path selection module and the congestion detection module are connected with the network monitoring module to obtain path selection parameters, and the optimal path selection for data stream transmission is carried out. The invention provides an intelligent congestion avoidance routing method for the data flow in the network, and effectively improves the performance of the network and the user service quality experience.

Description

Intelligent rerouting method and device based on congestion awareness in software defined network
Technical Field
The invention relates to the field of communication, in particular to an intelligent rerouting method and device based on congestion perception in a software defined network.
Background
With the continuous emergence of novel networks such as the internet of things, the mobile internet, the cloud computing network and the like, the streaming media data shows explosive growth, and the problem of network link congestion is increasingly serious. The distributed architecture of the conventional network lacks control over network resources and global information, and thus efficient utilization of network resources is difficult. The software defined network architecture breaks the design concept of the traditional network system, separates the centralized control plane from the distributed data plane, and can realize the global optimization and the centralized control of the control plane and the high-performance network forwarding capability. The software defined network architecture mainly comprises a data plane, a control plane and an application plane. The data plane device only has a forwarding function and does not have a control function. The control plane has global information and centralized control capability of the underlying network. The application plane is open to the user for the user to change network requirements and study innovations. The centralized management and control and the programmable interface of the software defined network bring convenience for network management.
In recent years, there has been some related research into solving the problem of network congestion under software defined network architectures. First, for network congestion evaluation, there are threshold methods and fuzzy evaluation methods based on switch port data. But lack an overall congestion status assessment of the data transmission path. Secondly, the congestion management mechanism based on the software defined network is mainly divided into two categories, namely a terminal side and a network side. The congestion management mechanism at the terminal side is mainly an improvement on the traditional TCP protocol, and the congestion condition of a link is temporarily relieved by adjusting a sending window and a receiving window of the terminal, wherein the congestion management mechanism comprises an SCCP protocol, an SDTCP protocol and the like. However, based on the adjustment of the network congestion at the terminal side, a large number of other links in the network may be in an idle state, and network resources may not be fully utilized. The network side congestion control method mainly reroutes data on a congestion path, routes a data flow at a congestion node to a path with a low network link utilization rate, and relieves the network congestion condition. The method comprises an ECMP routing method, a multi-alternative path method, a congestion node avoidance selection method and the like. The rerouting method can better utilize the network resources, but the determination of the quality of the rerouting path is a difficulty faced by the method.
With the development of artificial intelligence, research and application of a reinforcement learning method for solving the problem of network routing are available. The method mainly comprises a reinforcement learning routing method based on the QoS of the service. The reinforcement learning method can adjust the decision in real time according to the feedback of the network, and is more intelligent.
In summary, from the current research, the network congestion management mainly includes the following two problems to be solved:
(1) congestion awareness
The traditional segmented link congestion judgment is difficult to obtain the whole congestion effect of a data transmission path, and inconvenience is brought to optimal path selection.
(2) Congestion recovery mechanism
The traditional network side congestion recovery mechanism lacks the consideration of network resource optimization. The characteristics of network state change are ignored, selection of backup paths is too rigid, and the network effect brought by rerouting paths is not considered.
Disclosure of Invention
The invention aims to provide an intelligent rerouting method and device based on congestion perception in a software defined network.
In order to achieve the purpose, the invention adopts the following technical scheme:
the intelligent rerouting method based on congestion perception in the software defined network comprises the following steps:
(1) initializing a network;
(2) the network monitoring process comprises the following steps:
after the network initialization is completed, the network monitoring module of the control layer periodically collects and calculates the network state of the data layer, including: link used load ratio
Figure BDA0001628371390000021
Rate of change of link load
Figure BDA0001628371390000022
End-to-end Delay of a data transmission path, packet loss rate L oss of the path and throughput TH of the path, and a network state of a data layer is used as an evaluation parameter for calculating the path quality of the network;
when data stream is transmitted, a network monitoring module of a control layer counts end-to-end Delay of a data transmission path, packet loss rate L oss of the path and throughput TH of the path, and takes the end-to-end Delay of the data transmission path, packet loss rate L oss of the path and throughput TH of the path as alternative path QoS data, and the alternative path QoS data is used for calculating a feedback value of data transmission effect;
(3) and (3) a congestion detection process:
based on fuzzy system method, congestion detection module of control layer compares used load of link
Figure BDA0001628371390000031
And rate of change of link load
Figure BDA0001628371390000032
Fuzzy evaluation is carried out on the two indexes to obtain a fuzzy evaluation value rank of the congestion quality of the current linkijAccording to the average congestion degree of all links on the path, obtaining the fuzzy evaluation value rank of the congestion quality of the pathi；
(4) And (3) routing selection process:
after the congestion detection module of the control layer finishes the congestion quality evaluation of all the alternative paths, the path selection module and the congestion detection module are connected with the network monitoring module to obtain path selection parameters, and the optimal path selection for data stream transmission is carried out.
The invention is further improved in that the specific initialization process in the step (1) is as follows: the user access layer sends a data stream transmission request between a source node and a destination node, a control layer establishes connection with a data layer in a network, then a routing module of the control layer calculates K shortest paths for the data stream according to a shortest path algorithm to be used as an alternative path set for stream transmission, and finally, a shortest path is selected for a newly arrived stream request control layer and is issued to the data layer for stream transmission.
A further improvement of the present invention is that in step (2) the link used load ratio
Figure BDA0001628371390000033
Calculated by equation (1):
Figure BDA0001628371390000034
in the formula (1), the reaction mixture is,
Figure BDA0001628371390000035
for the time t the link is already loaded,
Figure BDA0001628371390000036
is the total link capacity at time t;
rate of change of link load
Figure BDA0001628371390000037
Calculated by equation (2):
Figure BDA0001628371390000038
in the formula (I), the compound is shown in the specification,
Figure BDA0001628371390000039
for the time t the link is already loaded,
Figure BDA00016283713900000310
the link is loaded with capacity for time t-1.
The invention is further improved in that the specific process of the step (3) is as follows:
(3a) the congestion detection module of the control layer is connected with the network monitoring module to obtain the path congestion quality evaluation parameters: link used load ratio
Figure BDA0001628371390000041
And rate of change of link load
Figure BDA0001628371390000042
(3b) Fuzzification: adopting link used load ratio membership to link used load ratio
Figure BDA0001628371390000043
Carrying out fuzzy evaluation; applying link load change rate membership to link load change rate
Figure BDA0001628371390000044
Carrying out fuzzy evaluation;
(3c) fuzzy rule table mapping: link used load ratio
Figure BDA0001628371390000045
And rate of change of link load
Figure BDA0001628371390000046
Mapping is carried out through a fuzzy rule table to obtain a link congestion degree RANKij；
(3d) And (3) deblurring: link congestion degree RANK for defined fuzzy output value by adopting maximum degree membership methodijDefuzzification is carried out to obtain an evaluation value rank of the link congestion qualityij；
(3e) Fuzzy evaluation of path congestion quality:
for all paths p between source node and destination nodei∈ P to evaluate the path congestion quality to obtain the average congestion quality evaluation value rank of all links on the pathi：
Figure BDA0001628371390000047
In formula (3), rankiRepresenting a pair of paths p between a source node and a destination nodeiCongestion quality evaluation value, rankij(m) represents a path piThe congestion quality evaluation value of the link passed in the mth order, N denotes a path piTotal number of all links passed on.
The invention is further improved in that the specific process of the step (4) is as follows:
(4a) the path selection module is connected with the network monitoring module to obtain QoS data of all alternative paths, namely end-to-end Delay of a data transmission path, packet loss rate L oss of the path and throughput TH of the path;
(4b) The path selection module is connected with the congestion detection module to obtain congestion quality evaluation values rank of all alternative pathsi；
(4c) Calculating a network long term revenue value from the alternative path QoS data Rwd;
(4d) calculating all alternative paths p between the source node and the destination node in the period according to the reinforcement learning methodiRespective comprehensive quality evaluation values Qi(P,pi) According to all alternative paths p in the periodiRespective comprehensive quality evaluation values Qi(P,pi) Updating the Q value table of the alternative path;
(4e) according to all alternative paths p in the periodiRespective comprehensive quality evaluation values Qi(P,pi) The path selection module determines all alternative paths p in the next periodiProbability of being selected pi (p)i) All alternative paths p for the next cycleiProbability of being selected pi (p)i) The calculation method of (2) is shown as formula (6);
Figure BDA0001628371390000051
wherein the content of the first and second substances,
Figure BDA0001628371390000052
in formulae (6) and (7):
π(pi) All alternative paths p for the next cycleiThe probability of being selected, K is the total number of paths between each pair of source nodes and destination nodes; tau isnIs a temperature parameter, the time varying parameter controlling the degree of randomness of selecting a certain path; t represents the convergence time; tau is0And τTA final temperature representing the initial temperature and time T;
(4f) the path selection module updates the path and issues a data transmission flow table to the data layer; and (3) detecting whether the transmission of the data stream is finished, if not, returning to the step (2), evaluating and updating the path quality in a new period, and if the transmission of the data is finished, ending the network monitoring.
A further improvement of the invention is that in step (4c) the network long term benefit value Rwd is calculated by the following formula:
Rwd＝ω1·TH+ω2/Delay+ω3/Loss (5)
wherein: omega1，ω2，ω3Is the proportion of each evaluation index in the feedback value.
The invention further improves the periodic alternative path comprehensive quality evaluation value Qi(P,pi) Calculated by the following formula:
Qi(P,pi)←α×ranki×{Rwd+γ×maxy∈PQi(P,y)}+(1-α)×Qi(P,pi) (4)
wherein, rankiIs the alternative path p of this periodiα is the learning rate, Qi(P,pi) Is the alternative path p of this periodiThe overall quality evaluation value of (2).
An intelligent rerouting device based on congestion perception in a software defined network comprises a user access layer, a data layer, a control layer and an application layer; the user access layer is composed of network users and used for making a data stream transmission request; the data layer is composed of a switch and a network link and realizes the forwarding function of the data stream; the control layer is composed of a software defined network controller, the software defined network controller comprises a network monitoring module, a congestion detection module and a routing module, the network monitoring module is used for periodically collecting link states in a network and end-to-end QoS data of each data transmission path in the network and transmitting the data to the congestion detection module and the routing module, the congestion detection module is used for carrying out fuzzy evaluation on the congestion degree of the data transmission path according to the data collected by the network monitoring module, the routing module is used for carrying out feedback according to a path congestion state evaluation value of the congestion detection module and path QoS data of the network monitoring module, the comprehensive quality of alternative paths is comprehensively evaluated by adopting a reinforcement learning method, and routing is carried out on the basis of the comprehensive quality of the alternative paths; the application layer is used for changing network requirements and research by users.
Compared with the prior art, the invention has the following beneficial effects: the method for evaluating the path congestion is a fuzzy system, and the method for selecting the path is a reinforcement learning method. Taking the link state of the data layer as a congestion quality evaluation parameter, and a congestion detection module of the control layer evaluates the path congestion quality; and taking the path congestion quality evaluation value and the path QoS data of the data layer as evaluation parameters, and carrying out comprehensive quality evaluation on the path by a path selection module of the control layer, selecting an optimal path with a certain probability based on the comprehensive evaluation value, and issuing the optimal path to the data layer for data stream forwarding. Firstly, by introducing a fuzzy system evaluation method, the invention realizes the comprehensive evaluation of the congestion degree of a data flow transmission path in the network, realizes the sensing capability of the network on the congestion path, reduces the dependence degree of the traditional congestion recovery mechanism on congestion detection, can pre-sense the occurrence of the path congestion condition in the network and make a response in advance; and the sensitivity of the network to the perception of the congestion path is improved to a certain extent, thereby improving the reliability of the network. Secondly, the method utilizes the QoS data of the data stream transmission path as an evaluation parameter and utilizes a reinforcement learning method to evaluate the comprehensive quality of the data transmission path, thereby realizing the intelligence of the selection of the data stream transmission path and improving the performance of the network and the experience of the user service quality.
Drawings
Fig. 1 is a structural diagram of the intelligent routing device.
Fig. 2 is a flowchart of the whole intelligent routing device.
Fig. 3 is a flow chart of a congestion detection module of the control layer.
Fig. 4 is a flow chart of a control layer routing module.
Fig. 5 is a graph of the congestion detection module link used load ratio membership function.
Fig. 6 is a graph of a membership function of a link load change rate of a congestion detection module.
Fig. 7 is a membership function diagram of the congestion quality evaluation value of the link of the congestion detection module.
In the figure, 101 is an application layer, 102 is a control layer, 103 is a data layer, and 104 is a user access layer.
Detailed Description
The invention will now be described in detail with reference to the drawings attached hereto and by way of examples.
In the examples of the present invention: firstly, a software-defined network communication environment is a necessary condition, fig. 1 is a concrete embodiment of the necessary condition, and fig. 2 is a data flow processing flow of fig. 1; second, the evaluation of the path congestion quality is one of the features of the present invention, fig. 3 is a flow of implementing the evaluation of the path congestion quality, and fig. 5, 6 and 7 are fuzzy relations of evaluation parameters in the process of evaluating the congestion quality; thirdly, the path selection method is the core of the invention, and fig. 4 is the flow of the path selection method.
The implementation environment of the present invention is divided into a user access layer 104, a data layer 103, a control layer 102, and an application layer 101. The user access layer 104 is composed of network users, and makes data stream transmission requests; the data layer 103 is composed of switches and network links, and realizes the forwarding function of data streams; the control layer 102 is composed of a software defined network controller, which mainly comprises a network monitoring module, a congestion detection module and a routing module; the application layer 101 is used for users to change network requirements and research innovations.
And the network monitoring module is used for periodically collecting link states in the network and end-to-end QoS data of each data transmission path in the network.
And the congestion detection module is used for carrying out fuzzy evaluation on the congestion degree of the data transmission path according to the data collected by the network monitoring module.
Fuzzification: and carrying out fuzzy evaluation on the link used load ratio and the link load change rate of all links on the alternative path based on the fuzzy relation in the invention.
Fuzzy rule mapping: fuzzy input values are input based on a rule table in the invention: link used load ratio, link load change rate and fuzzy output value: the link congestion level is mapped.
Defuzzification: and defuzzifying the link congestion degree based on a maximum membership method to obtain a link congestion quality evaluation value.
And (3) evaluating the congestion condition of the alternative path: and taking the average congestion quality of each link on the alternative path as the congestion quality evaluation value of the alternative path.
And the routing selection module is used for comprehensively evaluating the comprehensive quality of the alternative paths by applying a reinforcement learning method according to the path congestion quality evaluation value of the congestion detection module and the path QoS feedback data of the network monitoring module. And performs routing based on the value.
And calculating a long-term profit value of the path based on the path QoS feedback value of the network monitoring module.
And calculating the comprehensive quality evaluation value of the alternative path by applying a reinforcement learning method based on the congestion quality evaluation value of the alternative path and the long-term income value of the path.
And the routing module performs optimal routing on the data stream according to a defined formula and a certain probability.
The intelligent congestion avoidance routing method of the present invention is implemented in control layer 102. FIG. 1 is a diagram of the environment in which the present invention is implemented. The specific implementation process is as follows:
(1) network state initialization
As shown in fig. 1, in the network initialization phase, the user access layer 104 sends a data stream transmission request between a source node and a destination node, and the control layer 102 establishes a connection with the data layer 103 in the network. Next, the routing module of the control layer 102 calculates K shortest paths for the data stream according to a shortest path algorithm, and uses the K shortest paths as an alternative path set for stream transmission. Finally, the control layer 102 selects a shortest path for the newly arrived flow request, and sends the shortest path to the data layer 103 for flow transmission. The workflow of the network during the initialization phase is similar to that of a conventional network.
(2) Network monitoring process
Referring to fig. 2, after the network initialization phase is completed, but unlike the conventional network, the network monitoring module of the control layer 102 periodically collects the network status of the data layer 103 after the start of operation, which includes: link used load ratio
Figure BDA0001628371390000091
Rate of change of link load
Figure BDA0001628371390000092
The end-to-end Delay of the data transmission path, the packet loss rate L oss of the path, the throughput TH of the path, and the network state of the data layer 103 are used as evaluation parameters for calculating the path quality of the network.
Wherein if the link used load ratio at time t
Figure BDA0001628371390000093
Rate of change of link load
Figure BDA0001628371390000094
Calculated according to equations (1) and (2).
Figure BDA0001628371390000095
In the formula (1), the reaction mixture is,
Figure BDA0001628371390000096
for the time t the link is already loaded,
Figure BDA0001628371390000097
the total link capacity at time t.
If the link load change rate at time t
Figure BDA0001628371390000098
Comprises the following steps:
Figure BDA0001628371390000099
in the formula (I), the compound is shown in the specification,
Figure BDA00016283713900000910
for the time t the link is already loaded,
Figure BDA00016283713900000911
the link is loaded with capacity for time t-1.
The link load change rate describes a rate at which the link load at the present time increases relative to the link load at the previous time. It should be noted that: when the link load change rate is a positive value, the load on the link at the current moment is in an increasing state; when the load on the link is negative, the load on the link at the current moment is in a reduced state. The larger the absolute value of the link load change rate, the more significant the load change on the link.
Meanwhile, when data stream is transmitted, the network monitoring module of the control layer 102 counts end-to-end Delay, path packet loss rate L oss, and path throughput TH of the data transmission path of the data layer 103, and uses the value as alternative path QoS data, which is used for calculating a feedback value of data transmission effect.
(3) Congestion detection procedure
Firstly, the term 'fuzzy system' is introduced in the invention, and the fuzzy system refers to a method for carrying out decision management on a system by simulating human comprehensive reasoning.
Based on fuzzy system method, congestion detection module of control layer 102 compares load ratio of link used
Figure BDA0001628371390000101
And rate of change of link load
Figure BDA0001628371390000102
Fuzzy evaluation is carried out on the two indexes to obtain a fuzzy evaluation value rank of the congestion quality of the current linkijAccording to the average congestion degree of all links on the path, the fuzzy evaluation value rank of the congestion quality of the path can be obtainedi。
During data stream transmission, the congestion detection module of the control layer 102 periodically evaluates the congestion quality of all alternative paths between the source node and the destination node to obtain an alternative path piCongestion quality evaluation value rank ofi. The calculation process is shown in fig. 3, and specifically as follows:
(3a) the congestion detection module of the control layer 102 is connected to the network monitoring module to obtain a path congestion quality evaluation parameter: link used load ratio
Figure BDA0001628371390000103
And rate of change of link load
Figure BDA0001628371390000104
(3b) Fuzzification: referring to FIG. 5, the relationship of membership of the link utilization load ratio to the link utilization load ratio is used
Figure BDA0001628371390000105
Carrying out fuzzy evaluation; referring to FIG. 6, link load change rate is related to link load change rate by link load change rate membership
Figure BDA0001628371390000106
The blur evaluation was performed.
(3c) Fuzzy rule table mapping: referring to Table 1, the fuzzy rule table is defined with the "… if …" rule, e.g., the first "… if …" rule in the fuzzy rule table is defined as follows:
if the link has used the load ratio
Figure BDA0001628371390000107
Is low, the link load change rate
Figure BDA0001628371390000108
Is negative, then the link congestion degree RANKijIs perfect.
The fuzzy rule table defines 15 rules in total to represent the evaluation condition of the link congestion degree under different network environments, and the details are shown in table 1.
Fuzzy input variable link used load ratio
Figure BDA0001628371390000109
And rate of change of link load
Figure BDA00016283713900001010
Mapping is carried out through a fuzzy rule table to obtain fuzzy output value link congestion degree RANKij。
Table 1 fuzzy rule table for congestion detection module
Figure BDA00016283713900001011
Figure BDA0001628371390000111
(3d) And (3) deblurring: referring to fig. 7, the maximum membership method is adopted for the defined fuzzy output variable link congestion degree RANKijDefuzzification is carried out to obtain an evaluation value rank of the link congestion qualityij。
(3e) Fuzzy evaluation of path congestion quality:
for all paths p between source node and destination nodei∈ P, average congestion quality rating rank of all links on the pathiThe calculation method of the congestion quality evaluation value of the path is shown in the formula (3).
Figure BDA0001628371390000112
In formula (3), rankiRepresenting a pair of paths p between a source node and a destination nodeiCongestion quality evaluation value, rankij(m) represents a path piThe congestion quality evaluation value of the link passed in the mth order, N denotes a path piTotal number of all links passed on.
(4) Routing process
After the congestion detection module of the control layer 102 completes the congestion quality evaluation of all the alternative paths, the path selection module establishes connection with the congestion detection module and the network monitoring module to obtain path selection parameters, and performs optimal path selection for data stream transmission. The path selection module implementation process is shown in fig. 4, and specifically as follows:
(4a) the path selection module is connected with the network monitoring module to obtain QoS data of all the alternative paths, namely end-to-end Delay of a data transmission path, packet loss rate L oss of the path and throughput TH of the path.
(4b) Road surfaceThe path selection module is connected with the congestion detection module to obtain congestion quality evaluation values rank of all alternative pathsi。
(4c) And calculating a network long-term profit value Rwd according to the alternative path QoS data, wherein the calculation method is shown as a formula (5).
(4d) Calculating all alternative paths p in the period from the source node to the destination node by taking a reinforcement learning method as a calculation model according to the reinforcement learning methodiRespective comprehensive quality evaluation values Qi(P,pi) The calculation method is shown as the formula (4).
Qi(P,pi)←α×ranki×{Rwd+γ×maxy∈PQi(P,y)}+(1-α) (4)
×Qi(P,pi)
Wherein the content of the first and second substances,
Rwd＝ω1·TH+ω2/Delay+ω3/Loss (5)
in the formulae (4) and (5):
rwd this path p was selected at the previous momentiLong-term net path gain value, omega1，ω2，ω3Is the proportion of each evaluation index in the feedback value; rankiIs the alternative path p of this periodiα is the learning rate, Qi(P,pi) Is the alternative path p of this periodiThe overall quality evaluation value of (2).
(4e) The path selection module selects all the alternative paths p according to the periodiRespective comprehensive quality evaluation values Qi(P,pi) And updating the Q value table of the alternative path.
(4f) According to all alternative paths p according to the periodiRespective comprehensive quality evaluation values Qi(P,pi) The path selection module determines all alternative paths p in the next periodiProbability of being selected pi (p)i) All alternative paths p for the next cycleiProbability of being selected pi (p)i) The calculation method of (2) is shown in the formula (6).
Figure BDA0001628371390000131
Wherein the content of the first and second substances,
Figure BDA0001628371390000132
in formulae (6) and (7):
π(pi) All alternative paths p for the next cycleiThe probability of being selected, K is the total number of paths between each pair of source nodes and destination nodes; tau isnIs a temperature parameter, the time varying parameter controlling the degree of randomness of selecting a certain path; t represents the convergence time; tau is0And τTRepresenting the initial temperature and the final temperature for time T.
At the beginning of training τnThe method is large, and the optimal exploration on the alternative path is realized; and when the network environment tends to be stable, the value is smaller, and optimal convergence is realized.
(4g) The path selection module updates the path and issues a data transmission flow table to the data layer 103.
And (3) detecting whether the transmission of the data stream is finished, if not, returning to the step (2), evaluating and updating the path quality in a new period, and if the transmission of the data is finished, ending the network monitoring.
The software defined network controller comprises three modules, namely a network monitoring module, a congestion detection module and a routing module, wherein all the modules are realized in the controller. The network monitoring module periodically collects network link status and path end-to-end QoS data. Based on the fuzzy system method, the congestion detection module applies the data to carry out fuzzy evaluation on the congestion conditions of all alternative paths between each pair of source nodes and destination nodes in the network, so as to obtain the current congestion quality evaluation value of the alternative paths. Based on the current congestion quality evaluation value of the path and QoS feedback data of the path, the route selection module calculates a comprehensive evaluation value of the alternative path by applying a reinforcement learning method, and selects an optimal route for the data flow based on the value. The invention provides an intelligent congestion avoidance routing method for the data flow in the network, and effectively improves the performance of the network and the user service quality experience.
Through the above detailed description of the embodiments, the intelligent congestion avoidance path selection method and apparatus based on software defined network of the present invention, those skilled in the art can clearly understand that the network state can be obtained from the model, and the network congestion occurrence avoidance effect can be achieved; meanwhile, the path selection device for avoiding congestion in the software defined network provides support for the calculation method of the software defined network path, and comprises the steps of acquiring the data layer network state, evaluating the congestion degree of the data layer data transmission path and selecting the optimal path for data stream transmission.
The above embodiments are preferred examples of the present invention, and the present invention is not limited to the present embodiments, and includes sub-inventions within the scope of the features, methods, ideas and spirit thereof. The main features and ideas of the present invention and sub-inventions thereof are to be considered within the scope of protection.

Claims (6)

1. The intelligent rerouting method based on congestion awareness in the software defined network is characterized by comprising the following steps of:
(1) initializing a network;
(2) the network monitoring process comprises the following steps:
after the network initialization is completed, a network monitoring module of the control layer (102) periodically collects and calculates the network state of the data layer (103), including: link used load ratio
Figure FDA0002448633670000011
Rate of change of link load
Figure FDA0002448633670000012
End-to-end Delay of a data transmission path, packet loss rate L oss of the path, throughput TH of the path, and network state of a data layer (103) as evaluation parameters for calculating the path quality of the network;
when data stream is transmitted, a network monitoring module of a control layer (102) counts end-to-end Delay of a data transmission path, a packet loss rate L oss of the path and throughput TH of the path of a data layer (103), and takes the end-to-end Delay of the data transmission path, the packet loss rate L oss of the path and the throughput TH of the path as alternative path QoS data, wherein the alternative path QoS data is used for calculating a feedback value of data transmission effect;
(3) and (3) a congestion detection process:
based on fuzzy system method, congestion detection module of control layer (102) compares load ratio of link used
Figure FDA0002448633670000013
And rate of change of link load
Figure FDA0002448633670000014
Fuzzy evaluation is carried out on the two indexes to obtain a fuzzy evaluation value rank of the congestion quality of the current linkijAccording to the average congestion degree of all links on the path, obtaining the fuzzy evaluation value rank of the congestion quality of the pathi(ii) a The specific process is as follows:
(3a) the congestion detection module of the control layer (102) is connected with the network monitoring module to obtain the path congestion quality evaluation parameters: link used load ratio
Figure FDA0002448633670000015
And rate of change of link load
Figure FDA0002448633670000016
(3b) Fuzzification: adopting link used load ratio membership to link used load ratio
Figure FDA0002448633670000017
Carrying out fuzzy evaluation; applying link load change rate membership to link load change rate
Figure FDA0002448633670000018
Carrying out fuzzy evaluation;
(3c) fuzzy rule table mapping: link used load ratio
Figure FDA0002448633670000019
And rate of change of link load
Figure FDA00024486336700000110
Mapping is carried out through a fuzzy rule table to obtain a link congestion degree RANKij；
(3d) And (3) deblurring: link congestion degree RANK for defined fuzzy output value by adopting maximum degree membership methodijDefuzzification is carried out to obtain an evaluation value rank of the link congestion qualityij；
(3e) Fuzzy evaluation of path congestion quality:
for all paths p between source node and destination nodei∈ P to evaluate the path congestion quality to obtain the average congestion quality evaluation value rank of all links on the pathi：
Figure FDA0002448633670000021
In formula (3), rankiRepresenting a pair of paths p between a source node and a destination nodeiCongestion quality evaluation value, rankij(m) represents a path piThe congestion quality evaluation value of the link passed in the mth order, N denotes a path piTotal number of all links passed on;
(4) and (3) routing selection process:
after the congestion detection module of the control layer (102) finishes the congestion quality evaluation of all the alternative paths, the path selection module and the congestion detection module are connected with the network monitoring module to obtain path selection parameters and perform optimal path selection for data stream transmission;
the specific process is as follows:
(4a) the path selection module is connected with the network monitoring module to obtain QoS data of all alternative paths, namely end-to-end Delay of a data transmission path, packet loss rate L oss of the path and throughput TH of the path;
(4b) the path selection module is connected with the congestion detection module to obtain congestion quality evaluation values rank of all alternative pathsi；
(4c) Calculating a network long term revenue value from the alternative path QoS data Rwd;
(4d) calculating all alternative paths p between the source node and the destination node in the period according to the reinforcement learning methodiRespective comprehensive quality evaluation values Qi(P,pi) According to all alternative paths p in the periodiRespective comprehensive quality evaluation values Qi(P,pi) Updating the Q value table of the alternative path;
(4e) according to all alternative paths p in the periodiRespective comprehensive quality evaluation values Qi(P,pi) The path selection module determines all alternative paths p in the next periodiProbability of being selected pi (p)i) All alternative paths p for the next cycleiProbability of being selected pi (p)i) The calculation method of (2) is shown as formula (6);
Figure FDA0002448633670000031
wherein the content of the first and second substances,
Figure FDA0002448633670000032
in formulae (6) and (7):
π(pi) All alternative paths p for the next cycleiThe probability of being selected, K is the total number of paths between each pair of source nodes and destination nodes; tau isnIs a temperature parameter, and the time-varying parameter controls the random degree of selecting a certain path; t represents the convergence time; tau is0And τTA final temperature representing the initial temperature and time T;
(4f) the path selection module updates the path and issues a data transmission flow table to the data layer (103); and (3) detecting whether the transmission of the data stream is finished, if not, returning to the step (2), evaluating and updating the path quality in a new period, and if the transmission of the data is finished, ending the network monitoring.
2. The intelligent rerouting method based on congestion awareness in a software defined network as claimed in claim 1, wherein the specific process initialized in step (1) is as follows: a user access layer (104) sends a data stream transmission request between a source node and a destination node, a control layer (102) establishes connection with a data layer (103) in a network, then a routing module of the control layer (102) calculates K shortest paths for the data stream according to a shortest path algorithm to be used as an alternative path set for stream transmission, and finally, the control layer (102) selects a shortest path for a newly arrived stream request and sends the shortest path to the data layer (103) for stream transmission.
3. The intelligent rerouting method based on congestion awareness in software defined network as claimed in claim 1, wherein in step (2) the link used load ratio
Figure FDA0002448633670000033
Calculated by equation (1):
Figure FDA0002448633670000034
in the formula (1), the reaction mixture is,
Figure FDA0002448633670000035
for the time t the link is already loaded,
Figure FDA0002448633670000036
is the total link capacity at time t;
rate of change of link load
Figure FDA0002448633670000041
Calculated by equation (2):
Figure FDA0002448633670000042
in the formula (I), the compound is shown in the specification,for the time t the link is already loaded,
Figure FDA0002448633670000044
the link is loaded with capacity for time t-1.
4. The intelligent rerouting method based on congestion awareness in a software defined network as claimed in claim 1, wherein the network long term profit value Rwd in step (4c) is calculated by the following formula:
Rwd＝ω1·TH+ω2/Delay+ω3/Loss (5)
wherein: omega1，ω2，ω3Is the proportion of each evaluation index in the feedback value.
5. The intelligent rerouting method based on congestion awareness in software defined network as claimed in claim 4, wherein the candidate path comprehensive quality evaluation value Q in this periodi(P,pi) Calculated by the following formula:
Qi(P,pi)←α×ranki×{Rwd+γ×maxy∈PQi(P,y)}+(1-α)×Qi(P,pi) (4)
wherein, rankiIs the alternative path p of this periodiα is the learning rate, Qi(P,pi) Is the alternative path p of this periodiThe overall quality evaluation value of (2).
6. The apparatus for intelligent rerouting based on congestion awareness in a software defined network as claimed in claim 1, comprising a user access layer (104), a data layer (103), a control layer (102) and an application layer (101); the user access layer (104) is composed of network use users and is used for making a data stream transmission request; the data layer (103) is composed of a switch and a network link and realizes the forwarding function of the data stream; the control layer (102) is composed of a software defined network controller, the software defined network controller comprises a network monitoring module, a congestion detection module and a routing module, the network monitoring module is used for periodically collecting link states in a network and end-to-end QoS data of each data transmission path in the network and transmitting the data to the congestion detection module and the routing module, the congestion detection module is used for carrying out fuzzy evaluation on the congestion degree of the data transmission path according to the data collected by the network monitoring module, the routing module is used for carrying out feedback according to a path congestion state evaluation value of the congestion detection module and path QoS data of the network monitoring module, comprehensively evaluating the comprehensive quality of alternative paths by adopting a reinforced learning method and carrying out routing selection on the basis of the comprehensive quality of the alternative paths; the application layer (101) is used for user to change network requirements and research.
CN201810332577.0A 2018-04-13 2018-04-13 Intelligent rerouting method and device based on congestion awareness in software defined network Active CN108540384B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810332577.0A CN108540384B (en) 2018-04-13 2018-04-13 Intelligent rerouting method and device based on congestion awareness in software defined network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810332577.0A CN108540384B (en) 2018-04-13 2018-04-13 Intelligent rerouting method and device based on congestion awareness in software defined network

Publications (2)

Publication Number Publication Date
CN108540384A CN108540384A (en) 2018-09-14
CN108540384B true CN108540384B (en) 2020-07-28

Family

ID=63480545

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810332577.0A Active CN108540384B (en) 2018-04-13 2018-04-13 Intelligent rerouting method and device based on congestion awareness in software defined network

Country Status (1)

Country Link
CN (1) CN108540384B (en)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109039885B (en) * 2018-08-27 2022-01-07 广州猎游信息科技有限公司 Data transmission path selection method and device
CN109246231B (en) * 2018-09-29 2022-11-18 北京深度奇点科技有限公司 Intelligent routing method and intelligent routing equipment
CN109361601B (en) * 2018-10-31 2021-03-30 浙江工商大学 SDN route planning method based on reinforcement learning
CN109579861B (en) * 2018-12-10 2020-05-19 华中科技大学 Path navigation method and system based on reinforcement learning
US10979314B2 (en) 2019-01-24 2021-04-13 Vmware, Inc. Dynamic inter-cloud placement of virtual network functions for a slice
US10944647B2 (en) 2019-01-24 2021-03-09 Vmware, Inc. Dynamic inter-cloud placement of virtual network functions for a slice
US10958579B2 (en) 2019-05-14 2021-03-23 Vmware, Inc. Congestion avoidance in a slice-based network
US10892994B2 (en) 2019-05-14 2021-01-12 Vmware, Inc. Quality of service in virtual service networks
US11588733B2 (en) 2019-05-14 2023-02-21 Vmware, Inc. Slice-based routing
US11012288B2 (en) 2019-05-14 2021-05-18 Vmware, Inc. Congestion avoidance in a slice-based network
US10897423B2 (en) * 2019-05-14 2021-01-19 Vmware, Inc. Congestion avoidance in a slice-based network
CN110708242B (en) * 2019-10-28 2021-12-17 迈普通信技术股份有限公司 Traffic scheduling method and device, electronic equipment and storage medium
CN112383482B (en) * 2020-11-16 2021-10-08 北京邮电大学 Dynamic Q value route calculation method and device based on data plane
CN112671649A (en) * 2020-12-22 2021-04-16 广州技象科技有限公司 Path selection method and device based on Internet of things transmission fault detection

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104734954A (en) * 2015-03-27 2015-06-24 华为技术有限公司 Routing determination method and device used for software defined network (SDN)
CN105227481A (en) * 2015-09-02 2016-01-06 重庆邮电大学 The SDN congestion control method for routing of cost minimization is dispatched based on path cost and stream
CN106341346A (en) * 2016-09-08 2017-01-18 重庆邮电大学 Routing algorithm of guaranteeing QoS in data center network based on SDN

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9338096B2 (en) * 2013-08-30 2016-05-10 Futurewei Technologies, Inc. Multicast tree packing for multi-party video conferencing under SDN environment

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104734954A (en) * 2015-03-27 2015-06-24 华为技术有限公司 Routing determination method and device used for software defined network (SDN)
CN105227481A (en) * 2015-09-02 2016-01-06 重庆邮电大学 The SDN congestion control method for routing of cost minimization is dispatched based on path cost and stream
CN106341346A (en) * 2016-09-08 2017-01-18 重庆邮电大学 Routing algorithm of guaranteeing QoS in data center network based on SDN

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
A load balanced congestion aware routing mechanism for Software Defined Networks;Shadi Attarha;《2017 Iranian Conference on Electrical Engineering (ICEE)》;20170504;全文 *
Congestion Control Mechanism in Software Defined Networking by Traffic Rerouting;Akash Srikanth;《2018 Second International Conference on Computing Methodologies and Communication (ICCMC)》;20180216;全文 *
面向SDN的多路径调度算法研究;董晓林;《中国优秀硕士学位论文电子期刊》;20171231;全文 *

Also Published As

Publication number Publication date
CN108540384A (en) 2018-09-14

Similar Documents

Publication Publication Date Title
CN108540384B (en) Intelligent rerouting method and device based on congestion awareness in software defined network
CN112491714B (en) Intelligent QoS route optimization method and system based on deep reinforcement learning in SDN environment
CN113328938B (en) Network autonomous intelligent management and control method based on deep reinforcement learning
Xu et al. Experience-driven networking: A deep reinforcement learning based approach
CN108259367B (en) Service-aware flow strategy customization method based on software defined network
CN112437020B (en) Data center network load balancing method based on deep reinforcement learning
CN110557732B (en) Vehicle edge computing network task unloading load balancing system and balancing method
CN112486690B (en) Edge computing resource allocation method suitable for industrial Internet of things
CN112202672B (en) Network route forwarding method and system based on service quality requirement
CN108667734A (en) It is a kind of that the through street with LSTM neural networks is learnt by decision making algorithm based on Q
CN114884895B (en) Intelligent flow scheduling method based on deep reinforcement learning
CN114500360A (en) Network traffic scheduling method and system based on deep reinforcement learning
CN108512772A (en) Quality-of-service based data center's traffic scheduling method
CN111010341B (en) Overlay network routing decision method based on deep learning
CN111556173A (en) Service chain mapping method based on reinforcement learning
CN114567588B (en) Software defined network Qos routing method based on time delay prediction and double ant colony
CN113676407A (en) Deep learning driven flow optimization mechanism of communication network
CN106850859A (en) A kind of user's request distribution method of the CDN based on SDN
CN112867092A (en) Intelligent data routing method for mobile edge computing network
CN116170370B (en) SDN multipath routing method based on attention mechanism and deep reinforcement learning
CN116132361A (en) QoS and QoE-based network resource optimization adjustment method and system
CN115665207A (en) Method for improving data transmission speed in software defined internet of things
CN111555978B (en) SDN routing arrangement method with energy saving and service quality guarantee functions in smart grid
Liu et al. A routing model based on multiple-user requirements and the optimal solution
Mikhailenko et al. Analysis of the adaptive neural network router

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant